ORIGINAL RESEARCH ARTICLEpublished: 19 December 2011doi: 10.3389/fpsyg.2011.00353

Emerging signs of strong reciprocity in human ontogenyErin Robbins* and Philippe Rochat

Department of Psychology, Emory University, Atlanta, GA, USA

Edited by:

Steven E. Mock, University ofWaterloo, Canada

Reviewed by:

David Liu, University of California SanDiego, USAJeff Loucks, University ofWashington, USA

*Correspondence:

Erin Robbins, Emory Infant and ChildLab, Department of Psychology,Emory University, 36 Eagle Row,Atlanta, GA 30322, USA.e-mail: eerobbi@emory.edu

Strong reciprocity is considered here as the propensity to sacrifice resources to be kind orto punish in response to prior acts, a behavior not simply reducible to self-interest and alikely force behind human cooperation and sociality.The aim was to capture emerging signsof strong reciprocity in human ontogeny and across highly contrasted cultures. Three- and5-year-old middle class American children (N = 162) were tested in a simple, multiple round,three-way sharing game involving the child, a generous puppet, and a stingy puppet. At theend of the game, the child was offered an opportunity to sacrifice some of her personalgains to punish one of the puppets. By 3 years, American children demonstrate a willing-ness to engage in costly punishment. However, only 5-year-olds show some evidence ofstrong reciprocity by orienting their punishment systematically toward the stingy puppet.Further analyses and three additional control conditions demonstrate that such propensityis not simply reducible to (a) straight imitation, or (b) inequity aversion. To assess the rela-tive universality of such development, a group of 5- to 6-year-old children from rural Samoa(N = 14) were tested and compared to age and gender-matched American children. Samoanchildren did not manifest the same propensity toward strong reciprocity. The results areinterpreted as pointing to (1) the developmental emergence of an ethical stance between3 and 5 years of age, and (2) that the expression of such stance by young children coulddepend on culture.

Keywords: strong reciprocity, costly punishment, cross-cultural psychology, sharing, egalitarianism, moral

reasoning

INTRODUCTIONDefectors of potential cooperation tend to be punished, even inanonymous interactions and even when such punishment is costly,not rational in either long or short economic terms, with the caveatthat it can vary across cultures (Henrich et al., 2006; Henrichand Henrich, 2007). Strong reciprocity captures the propensityto sacrifice resources, either to reward or to punish, in response toperceived kindness or un-kindness (Fehr et al., 2002). The ques-tion of interest here is when does such propensity start to emerge inhuman ontogeny and what is its significance in child development.

Recent evidence suggests that human pro-social tendencies aredeeply rooted in development, already present from the mid-dle of the first year of life, as in the case of infants detectingpro-social versus anti-social acts perpetrated by third-party pro-tagonists (Kuhlmeier et al., 2003; Hamlin et al., 2007). By theirsecond birthday, children manifest the explicit inclination to helpand collaborate with others (Warneken et al., 2007; Tomassello,2008), as well as to display empathic responses toward sufferingothers (Zahn-Waxler et al., 1992; Eisenberg and Fabes, 1998). It isalso by the second year of life that children begin to show explicitattention to social norms (Kagan, 1988).

Drawing on an analogy with existing theories on human lan-guage (Pinker, 2008), some theories postulate the existence ofa universal and innate moral instinct of the species, a moralgrammar (Hauser, 2006) unique to humans and deeply rootedin the evolution of particular emotional responses (Haidt andJoseph, 2004). However, the analogy with language remains

problematic (Dupoux and Jacob, 2007), particularly in view ofchildren’s slow developing sense of fairness or egalitarianism (i.e.,inequity aversion, Fehr et al., 2008) and moral reasoning (Piaget,1932; Kohlberg, 1981), both showing significant variations acrosscultures (Snarey, 1985; Rochat et al., 2009).

By 21 months children are explicit about possession. Personalpronouns and adjectives like “mine!” become part of the child’svocabulary as an index of explicit ownership over things, openingup possibilities for sharing, negotiation, and bartering with others(Faigenbaum, 2005). Research documents an important devel-opmental progression in the sharing of rewards and resources.Three-year-olds demonstrate marked self-maximizing (selfish)and self-interested tendencies when asked to share, as in the exam-ple of distributive justice games (Lane and Coon, 1972; Hook,1978). Cross-cultural studies show that self-maximizing tenden-cies by 3-year-olds tend to be moderated by the social and physicalenvironment of the child. Self-interest is significantly attenuated in3-year-olds growing up in rural, more communal small-scale tra-ditional societies of Fiji and Peru when compared to middle classchildren from urban and industrial Brazil, China, or United States(Rochat et al., 2009). However, despite these variations, it appearsthat regardless of culture, there is a decrease in self-maximizingtendencies in sharing between 3 and 5 year of age (Rochat et al.,2009). By 5 years, children are reported to apply a strict equityor egalitarianism principle to their distributions. Sharing and dis-tributive justice are presumed to develop with the child’s abilityto reason in terms of proportion (Inhelder and Piaget, 1964),

www.frontiersin.org December 2011 | Volume 2 | Article 353 | 1

Robbins and Rochat Emerging signs of strong reciprocity

eventually leading them to apply principles of proportionality byfactoring relative work, effort, production, need, or starting wealthfrom around 7 years of age (Damon, 1975).

At an age when children might still be overwhelmingly driven byself-maximizing tendencies, they also appear capable of generatingsocial preferences and value judgments about others on the basisof how they give. From 4 years of age, children base social judg-ments on absolute quantities of a distributed good, for examplejudging one of two puppets as “nicer” because it gives them morevaluable chips (McCrink et al., 2009). By 5 years, children begin toshow signs that they judge the relative niceness of the puppet basedon the proportion of the chips the puppet is willing to sacrifice, inthe same way that adults tend to judge whether a resource distri-bution is more or less equitable based on proportional reasoning(McCrink et al., 2009).

Recent evidence also suggests that by 3.5 years of age, evenwhen they are not themselves recipient of a distribution, childrendemonstrate selective preference in their distribution of resourcesbased on past relations (Olson and Spelke, 2008). In their study,Olson and Spelke asked 3- to 4-year-olds to help a doll distrib-ute resources to other dolls that were described as either closerelations (in-group) or strangers (out-group); dolls that had orhad not previously shared with the distributing doll (direct reci-procity); or dolls that did or did not previously share with otherdolls (indirect reciprocity). In this situation children demonstratedtendencies suggestive of an early sensitivity to close relations as wellas direct reciprocity (i.e., tit-for-tat exchanges) and indirect reci-procity (long term pay-off/reputation based exchanges). Note thatall this is demonstrated even though the sharing game involvespuppet dolls and not real protagonists.

Although tendencies toward direct and indirect reciprocity areboth considered as foundational aspects of cooperation and appearto“extend deep in human development”(Olson and Spelke, 2008),it is not clear yet how such tendencies eventually develop to becomestrong reciprocity, which represents a marked qualitative shift inmeaning. Strong reciprocity, as indexed by costly punishment,entails the sacrifice material resources to punish another individ-ual and is considered here as the expression of principled decisionsby the child. Engaging in costly punishment therefore requires thechild to forego personal gain to enforce a norm (e.g., for fairness)that has been violated. In the willingness to sacrificing her ownresources, the child demonstrates an ethical (principled) stancetoward others that is qualitatively different and presumably moreadvanced than simple reciprocity (e.g., tit-for-tat reciprocation).In addition, the foundational aspect of strong reciprocity needs tobe specified. For example, is it based on the adoption of some moralprinciple (what we call here an ethical stance), or might it restonly on an early propensity toward inequity aversion as proposedby Fehr et al. (2008)? Furthermore, little is known regarding theinfluence of culture on the development leading children from areciprocity that might be direct or indirect to one that is strong, i.e.,principled and presumably independent of future self-maximizingrewards.

In this context, the question we asked is twofold: First, whatare the ontogenetic origins of strong reciprocity and second, whatmight be the potential influence of culture in the expression ofsuch a tendency in children’s development?

As a general working hypothesis, we expected that strongreciprocity (as expressed in selective costly punishment) woulddevelop in the preschool years, becoming evident by 5–6 yearswhen children are also documented as becoming proficient andcomplex readers of others’ minds (Wellman and Liu, 2004), adevelopment that is predictable across cultures (Callaghan et al.,2005).

In a series of studies, we present evidence that first signs ofstrong reciprocity emerge by 5 years. However, we also providesome evidence that contrary to our expectations, the expressionof strong reciprocity by 5- to 6-year-olds can vary depending onculture (i.e., affluent, urban United States versus rural, traditional,and group-oriented Samoa, see Study 6).

GENERAL METHODOLOGYIn six different studies, children participated in a triadic (three-way), multi-round sharing game in which they were asked to splitcoins (poker chips of uniform size, shape, and color) betweenthemselves and two protagonists. To provide an incentive for play-ing, prior to the onset of the game children visited a makeshift toystore where the experimenter explained how accumulated coinscould be used to purchase various prizes (e.g., small toys andstickers <$1 USD). Children were also informed that their shar-ing partners could similarly spend their winnings. To ensure thatchildren had associated the coins with the ability to win prizes,at several points throughout the game children were prompted toexplain how the coins worked.

Study 1 examined children’s sense of reciprocity by asking par-ticipants to share coins with stingy and generous partners. Study 2replicated the task and then expanded this methodology to includea more direct test of strong reciprocity, a costly punishment taskin which children could sacrifice material gains to punish one(or both) of the sharing partners. Three controls (Studies 3–5)assessed whether children were sensitive to the actions of the stingyor generous puppets. Finally, based on the same procedure, a com-parison with a group of same age Samoan children in Study 6offered the opportunity to examine potential cross-cultural vari-ations in the emerging expression of strong reciprocity (selectivecostly punishment).

BASELINE ASSESSMENTIn order to assess children’s general sense of equitable and propor-tional sharing, a pre-test was performed first. Children were askedto split nine coins between three identical toy ducks. We limitedthe number of items shared to nine to avoid an exclusive relianceon counting, which was presumably not yet in the register of theyounger children.

For the pre-test trial and the four sharing rounds that followed(see description of the triadic sharing game below), an unfamiliarfemale Experimenter sat at a table opposite the child and arrangedthe valuables to be split (poker chips) in a circular formation onthe center of the table before instructing1 children to “split thecoins” (Figure 1). In the pre-test and in the triadic sharing gamethat followed, children indicated ownership of the coins by placing

1“Split” was used in place of other instructions (e.g., “share” or “divide”) which areoften connotative of parsing a given object equitably, or into equal parts.

Frontiers in Psychology | Developmental Psychology December 2011 | Volume 2 | Article 353 | 2

Robbins and Rochat Emerging signs of strong reciprocity

FIGURE 1 | Experimental situation in the United States. Identicalpuppets were animated in turn by the Experimenter during the three-waysharing game (here the “stingy” puppet maximizes coins for himself).

them on the table in front of the intended recipient. After thechild confirmed this distribution, the winnings were depositedinto banks (transparent plastic containers, each belonging to oneof the sharing protagonists) for safekeeping.

TRIADIC SHARING GAMEFollowing baseline pre-test, the child was tested in a triadic (three-way) sharing game using nine pretend, yet valuable coins (pokerchips). In four successive trials, the child was asked to split thecoins in turn with two puppets that acted either with marked gen-erosity or marked stinginess (Studies 1, 2, and 5). In three controlconditions, children participated in the same triadic share task, butwith a slight modification. In Study 3 (Generous Control) childreninteracted with two generous puppets. In Study 4 (Stingy Control)children interacted with two stingy puppets. Finally, in Study 5(Non-agentive Control), children interacted with two passive pup-pets. In this latter control condition, children are presented witha pre-established distribution without seeing the puppets activelysharing either stingily or generously.

The two identical,high-quality plush hand-puppets (43 cm tall)dressed as kings sat at the table to the right and left of the child andwere animated by the Experimenter with the same vocalizationsand mannerisms throughout the experiment (Figure 1). In foursuccessive rounds, the two puppets always started, one sharinggenerously (four coins to the other puppet, four coins to the child,and only one to itself), and one sharing stingily (keeping sevencoins for itself, giving one to the other puppet, and only one to thechild). By sharing last, children observed the maximum effect ofthe puppets’ generous or stingy behaviors.

In all studies, when it was the child’s turn, we recorded and ana-lyzed as a dependent measure the total number of coins distributedby the child to herself and to the two puppets as a function of the foursharing rounds.

Researchers counterbalanced both the position of the charac-ters (i.e., the location of the generous and stingy puppet relativeto the child) as well as which puppet shared first to avoid the

potential of a side bias or order effect. Children were clearly andunambiguously informed that each protagonist would keep andcould accumulate received coins to purchase attractive toys at amakeshift toy store they visited prior to testing. To re-emphasizethe accumulative nature of the game, at the end of each round theExperimenter asked the child to compare her bank with the banksof the two puppets and determine which player had accumulatedthe most coins.

SELECTIVE COSTLY PUNISHMENT TESTIn Studies 2–6, at the conclusion of the triadic sharing game chil-dren were probed for signs of strong reciprocity in a final costlypunishment test. After the last round of sharing, the Experimenterinstructed children to examine the contents of the three banksand determine which player had the most coins. Next, the Exper-imenter asked whether the outcome of the game was fair. It wasthen proposed that as the “banker” of the game, the Experimentercould take five coins away from a puppet of the child’s choosing,but only if the child gave up one of her own coins. To underscorethe costly nature of this decision, the Experimenter reminded chil-dren that coins sacrificed by the child or taken from the puppetscould not be used in the store. The Experimenter then asked ifthe child would like to proceed. A child who responded “yes” hadto give the Experimenter one of her own coins before identifyingthe puppet that should be punished2. The Experimenter offered achance for the child to engage in costly punishment again by ask-ing if she would like to sacrifice another coin (i.e., “Would you liketo give me another coin and I can take five coins away from oneof the kings?”). This procedure repeated until the child declinedto offer a coin, or until one of the banks was emptied. For thedependent measure, we recorded the number of coins sacrificed topunish each puppet.

PARTICIPANTSExcept where noted in Study 6 (cross-cultural comparison), wesampled children from predominantly middle- to upper-middleclass families living in metro-Atlanta, GA, USA (pop: >5 mil-lion). In accordance with our IRB-approved protocol, childrenwere enrolled into the study through the consent of parents, whowere invited by phone and e-mail to participate. Half of the partici-pants completed the study at a university research lab. We recruitedother children from small after-school programs and pre-schoolsless than 3 km from the university campus. When appropriate, weobtained additional consent from teachers and school administra-tors. Table 1 provides descriptive statistics of the participants inStudies 1–6.

STUDY 1 (CHILD–STINGY–GENEROUS TRIADIC SHAREWITHOUT COSTLY PUNISHMENT)HYPOTHESESOn the basis of prior studies that have found age-related trendsin egalitarian sharing behavior (Fehr et al., 2008; Rochat et al.,2009), we hypothesized that 3- and 5-year-olds would differ in

2The word “punishment” was not used to explain this task to participants in aneffort to avoid biasing children toward participating.

www.frontiersin.org December 2011 | Volume 2 | Article 353 | 3

Robbins and Rochat Emerging signs of strong reciprocity

Table 1 | Descriptive statistics of participants in Studies 1–6.

Males (N ) Females (N ) Mean age ± SD

(in months)

Study 1 (Child–Stingy–GenerousTriadic Share without

Costly Punishment), N = 36a

3-year-olds (N = 18) 10 8 44.27 ± 4.85

5-year-olds (N = 18) 11 7 67.33 ± 5.85

Study 2 (Child–Stingy–GenerousTriadic Share and Costly

Punishment), N = 66

3-year-olds (N = 33) 18 15 42.95 ± 4.89

5-year-olds (N = 33) 18 15 64.15 ± 4.90

Study 3 (Generous Control withTriadic Share and Costly

Punishment), N = 24

3-year-olds (N = 12) 4 8 44.25 ± 6.03

5-year-olds (N = 12) 4 8 65.89 ± 7.68

Study 4 (Stingy Control withTriadic Share and Costly Punishment),

N = 243-year-olds (N = 12) 7 5 45.00 ± 4.35

5-year-olds (N = 12) 5 7 60.92 ± 5.90

Study 5 (Non-Agentive Control withTriadic Share and Costly

Punishment), N = 18b

5-year-olds (N = 18) 9 9 64.17 ± 3.62

Study 6 (Cross-Cultural Comparison of us and Samoan 5-Year-Old

Children), N = 28c

Samoa (N = 14) 7 7 68.30 ± 5.83

US (N = 14) 7 7 65.18 ± 5.75

For 3-year-olds in all studies, participants ranged between 34 and 50 months and

5-year-olds ranged between 58 and 70 months. aStudy 1: two participants were

omitted on the basis that they did not complete the study, yielding an attrition

rate of 6%. bTo control for the possibility that inequity aversion was driving the

selective costly punishment results amongst older children, Study 5 sampled only

5-year-olds. cThe US children included in Study 6 were randomly selected from

the sample of 5-year-olds who participated in Study 2.

their distribution of the coins, particularly with regard to the stingyand generous puppets. As for rationale, we considered that if reci-procity guides children’s sharing, they should give more coins tothe generous rather than the stingy puppet (who violates an expec-tation of fairness) when it is their turn to split. We hypothesizedthat this tendency should become more marked as a function ofthe four sharing rounds as children observe the consistency of thegenerous and stingy puppets’ actions, and that this expression ofreciprocity should be evident by 5 years of age and not earlier.

RESULTSWe analyzed the number of coins (out of nine) distributed toeach protagonist as the dependent measure in a 4 (round) × 3(recipient) × 2 (age) × 2 (gender) mixed-design ANOVA. Chil-dren’s performance on the three-way sharing pre-test, coded asa binomial variable indicating whether or not children split coinsequally (3–3–3) between the three dolls, was included as covariateto control for potential differences in children’s abilities to reasonproportionally. Neither gender nor pre-test performances werefound to be significant factors, F 1,31 = 1.57 and 0.622, respectively.

Analyses yielded a three-way interaction of round, recipi-ent, and age, F 6,29 = 2.57, p = 0.040, η2 = 0.080. Taking into

consideration only the child as recipient, in a follow-up testto assess children’s self-maximizing tendencies over the courseof the sharing game we observed a significant main effect ofround, F 3,32 = 3.41, p = 0.029, η2 = 0.179. Overall, as a func-tion of round, children of both ages tended to accumulate morecoins for themselves. However, note that there was a marginallysignificant interaction of round by age (F 3,32 = 2.44, p = 0.082,η2 = 0.078) suggesting a stronger trend for younger children.The mean difference in the number of coins distributed tothe child between Rounds 1 and 4 was greater for 3-year-olds(MD ± SE = 2.66 ± 0.918, p < 0.01 based on Bonferroni-adjustedcomparisons) and not statistically significant for 5-year-olds(MD ± SD = 0.301 ± 0.664).

Additional follow-up tests assessed children’s treatment of thetwo puppets across the four rounds of sharing. Amongst 3-year-olds we observed no significant main effects or interactionsof round and recipient. On average, these children distributedroughly the same number of coins to the stingy (M ± SD =2.05 ± 1.60) and generous (M ± SE = 1.98 ± 1.76) puppet ineach round (see Figure 2, left panel). Therefore, 3-year-oldsdid not express any selective discrimination between the twoprotagonists. In contrast, 5-year-olds increasingly discriminatedbetween stingy and generous characters as a function of rounds(see Figure 2, right panel). There was a significant inter-action of round by recipient for 5-year-olds (F 3,15 = 2.91,p = 0.037, η2 = 0.084). Although the mean difference in thenumber of coins given to the generous and stingy was notsignificant in Round 1 (MD ± SE = 0.048 ± 0.387), childrensignificantly favored the generous over the stingy puppet inRound 2 (MD ± SE = 0.667 ± 0.226), Round 3 (MD ± SE =0.952 ± 0.337), and Round 4 (MD ± SE = 1.10 ± 0.388), allp < 0.05 based on pairwise comparisons, Bonferroni-adjusted).

Together, the results of the triadic sharing game indicatedthat sensitivity to the relative generosity or stinginess of thesharing partners, as a function of round, is evident by 5-year-oldsonly. Five-year-old children in our sample maximized their gains(although to a lesser extent than their 3-year-old counterparts)while simultaneously becoming more discriminating in their dis-tributions to the stingy over the generous puppet. These findingsindicate that 5- but not 3-year-olds deal with sharing partnersdepending on their character (generous or not) in acts of reci-procity and in response to acts that defy norms of fairness. How-ever, without evidence of self-sacrifice and costly punishment onthe part of the child, it is difficult to determine whether chil-dren are able to by-pass mere reciprocity to also show signs ofstrong reciprocity. The next study was devised to test for thispossibility.

STUDY 2 (CHILD–STINGY–GENEROUS TRIADIC SHARE WITHCOSTLY PUNISHMENT)In a more direct test of strong reciprocity, we repeated with 66children (33 three-year-olds and 33 five-year-olds, see Table 1) theTriadic Sharing game described in the first study, adding a follow-up condition (the Costly Punishment task) to determine whetherchildren would be willing to sacrifice some of their own accumu-lated coins to punish one of the two protagonists, particularly thestingy, unfair puppet.

Frontiers in Psychology | Developmental Psychology December 2011 | Volume 2 | Article 353 | 4

Robbins and Rochat Emerging signs of strong reciprocity

FIGURE 2 | Children’s distributions in triadic sharing in Study 1.

Vertical axis represents mean number of coins (maximum nine)distributed by 3-year-olds (left panel) and 5-year-olds (right panel) to each

protagonist (child, stingy, and generous) as a function of round (1–4). Thehorizontal line represents absolute equity (three coins). Bars represent±1 SE.

HYPOTHESESWe expected to replicate the findings in Study 1 regarding the fourrounds of three-way sharing with the stingy and generous puppets.We predicted that 5-year-olds would show reciprocity comparedto 3-year-olds who should self-maximize more and be less selectiveregarding the stingy or generous character of the puppets. In rela-tion to strong reciprocity, in the additional costly punishment taskwe expected 5-year-olds to engage in costly sacrifice to punish thestingy protagonist when offered the opportunity, going above andbeyond reciprocity, and thus starting to take a principled ethicalstance (see introduction for developmental rationale).

RESULTSAs in Study 1, we analyzed the number of coins (out of 9) distrib-uted to each player as the dependent measure in a 4 (round) × 3(recipient) × 2 (age) × 2 (gender) mixed-design ANOVA. Neithergender (F 1,61 = 0.628) nor children’s performance on the three-way pre-test, included here as a covariate (F 1,61 = 0.446), werefound to be significant factors.

Results yielded a significant three-way interaction of round,recipient, and age, F 6,59 = 3.73, p < 0.01, η2 = 0.531. In afollow-up test to assess children’s degree of self-maximizing andconsidering only what children gave to themselves, we observeda main effect of round, F 3,62 = 7.57, p < 0.01, η2 = 0.544. On

average, both 3- and 5-year-olds increased the number of coinsgiven to themselves over the course of the game, from Round 1(M ± SD = 3.97 ± 2.11) to Round 4 (M ± SD = 4.91 ± 2.26).

With regard to their treatment of the two puppets, overthe four sharing rounds we found no significant trends for 3-year-olds. These younger children did not differentiate signif-icantly between the stingy (M ± SD = 2.24 ± 1.50) and gener-ous (M ± SD = 2.20 ± 1.58) puppets at any point in the game(Figure 3, left panel). In contrast, as in Study 1, 5-year-oldsdiscriminated between the puppets. For these children, a signif-icant interaction of round and recipient (F 3,20 = 5.65, p < 0.01,η2 = 0.361) points to a preferential treatment of the generous overthe stingy protagonist. As predicted, the mean difference in thenumber of coins distributed to the generous and stingy puppetwas significant in Round 2 (MD ± SE = 0.635 ± 0.260), Round3 (MD ± SE = 0.458 ± 0.254), and Round 4 (MD ± SE = 0.667 ±0.278), all p < 0.01 based on Bonferroni-adjusted pairwisecomparisons. This analysis of the recipient by round interactionsuggests that as the game progressed, 5-year-olds gave more tothemselves at the expense of the stingy rather than the generouspuppet. The disparity between puppets (particularly in Rounds 2–4, see Figure 3, right panel) is likely a result of the child penalizingthe stingy protagonist by giving more to herself rather than thechild sacrificing her own coins to give more to the generous puppet.

www.frontiersin.org December 2011 | Volume 2 | Article 353 | 5

Robbins and Rochat Emerging signs of strong reciprocity

FIGURE 3 | Children’s distributions in triadic sharing in Study 2.

Vertical axis represents mean number of coins (maximum nine)distributed by 3-year-olds (left panel) and 5-year-olds (right panel) to each

protagonist (child, stingy, and generous) as a function of round (1–4). Thehorizontal line represents absolute equity (three coins). Bars represent±1 SE.

In the follow-up costly punishment task, the inclination toengage in costly punishment did not differ between age groups:76% of 3-year-olds (N = 25) and 91% of 5-year-olds (N = 30)opted to punish at least once. Because the puppets differed withregard to their stingy or generous acts, we hypothesized that fair-minded children should be less inclined to equally punish bothpuppets. A two-tailed Fisher’s exact test compared the percentageof children in each age group who punished in successive attemptsboth puppets. Sixty-four-percent of 3-year-olds did so, comparedto only 17% of 5-year-olds, p < 0.01.

To further investigate the selective orientation of children’scostly punishment as a speculative expression of strong reciprocity,we used a linear regression model with the child’s age (in months)to predict (out of all coins sacrificed) the percentage of coins givenup to punish the stingy puppet. We observed a significant, moder-ate to strong association ( R_1, 552 = 0.320, p < 0.01) between ageand punishment orientation. These data suggest that as a functionof age, when children sacrifice coins, they are increasingly selectivein orienting their punishment toward the stingy puppet. Note thatby 60 months, fewer children punish both puppets and no childrenpunish the generous puppet only (Figure 5, left panel).

Collectively, the results of the costly punishment task indicatethat unlike 3-year-olds, 5-year-olds selectively orient punishmenttoward the stingy character. We interpret this as first signs of strongreciprocity.

ALTERNATIVE HYPOTHESES AND FURTHER ANALYSESRegarding the triadic sharing game, the differing allocations by thechild to the two protagonists could rest on straight imitation (e.g.,copying identically the actions of the stingy or generous puppet).Children might also focus on the overall unequal accumulation ofcoins in the banks that were transparent, hence public. Because thegenerous puppet always receives the least amount of coins prior tothe child’s turn, the number of coins in this bank is typically lessthan that of the other two protagonists. It is therefore possible thatchildren gave more coins to the generous puppet in order to rec-tify this perceptual inequity (inequity aversion hypothesis), ratherthan as a means of expressing approval or spite for the puppet’sgenerous versus stingy actions (ethical stance hypothesis).

To test these two alternative accounts, we identified 13 poten-tial ways for children to distribute coins during the triadic share.To probe the alternative imitation account, we analyzed the extent

Frontiers in Psychology | Developmental Psychology December 2011 | Volume 2 | Article 353 | 6

Robbins and Rochat Emerging signs of strong reciprocity

to which children imitated the stingy and generous puppet in astrict “tit-for-tat” manner (i.e., giving to each of puppet the exactnumber of coins that the child’s received from each) during anyof four rounds of triadic sharing. We found that such emulationof either puppet was evident only for a small minority of childrenat both ages (22% of 3-year-olds and only 5% of 5-year-olds).There was no significant age difference in this trend based onchi-square calculations (p = 0.143). We therefore conclude thatstraight imitation alternative cannot account for our results.

To probe the inequity aversion hypothesis we reasoned that ifchildren are responding based on a perceived perceptual inequity,then the most common sharing strategies should be those inwhich children favor the generous puppet to level the numberof coins in the three banks (generous-oriented sharing). Chil-dren were only coded as engaging in generous-oriented sharing ifthey gave systematically more to the generous compared to eitherthe self or the stingy puppet. Self-maximizing or stingy orientedstrategies were therefore incompatible with such sharing (mutuallyexclusive). Analysis reveals that generous-oriented strategies wereuncommon, with 13% 3-year-olds and only 8% of 5-year-oldssharing this way3. The inequity aversion hypothesis does not seemto hold.

Finally, in relation to costly punishment, we analyzed whetherchildren engaged in costly sacrifice not to punish proper, but ratherto re-establish equal distribution among the protagonists (inequityaversion hypothesis). We calculated an index of inequity by com-puting for each participant the proportion of coins in each bank atthe end of the four round distributions (total accumulated coins instingy’s bank/(total accumulation of coins in child’s bank + totalaccumulated coins in generous’s bank). As a result, and consideringmathematical lower and upper limits, the index of inequity couldvary between a minimum value of 0.68 and a maximum valueof 2.86, with greater values indicating greater levels of inequitybenefiting the stingy puppet (maximum possible accumulationfor the stingy puppet which was 80 coins). We reasoned that theinequity aversion hypothesis would be supported if the index ofinequity (as it emerges at the end of the four rounds of distrib-ution and depending only on the child’s distribution) would (1)positively correlate with each child’s inclination to engage in costlypunishment and (2) would positively correlate with each child’sinclination to punish the doll that typically ended up with themost coins (i.e., the stingy puppet).

Non-parametric correlation test between our index of inequityand whether or not the child opted toward costly punishmentyielded non-significant results for both age groups, respectivelyrs31 = −0.15 for 3-year-olds and rs31 = −0.25 for 5-year-olds.Parametric correlation tests between index of inequity and theproportion of all coins sacrificed to punish the stingy puppet, alsoyielded non-significant results for both age groups, respectivelyr31 = 0.018 for 3-year-olds and r31 = 0.287 for 5-year-olds. Fromthese results, we conclude that the inequity aversion hypothesiscannot readily account for our costly punishment results.

3Percentages for the sharing patterns in Study 1 were comparable and did not differsignificantly from those reported here for Study 2.

To assert that children are indeed more or less sensitive tothe stingy or generous character of the puppets, we performedthree control experiments that are presented next. Specifically, werepeated the triadic share and costly punishment task with threenew cohorts of children sharing in turn with either (a) two iden-tical generous puppets (Study 3), (b) two identical stingy puppets(Study 4), or (c) puppets that were non-agentive in the splitting ofcoins (Study 5).

STUDY 3 (GENEROUS CONTROL TRIADIC SHARE ANDCOSTLY PUNISHMENT)To determine whether children are sensitive to the generous char-acter of the puppet, we ran a first control sampling 24 children(12 three-year-olds and 12 five-year-olds, see Table 1) who playedthe Triadic Sharing game and Costly Punishment task describedpreviously, but with identically generous puppets.

HYPOTHESESRegarding the four rounds of three-way sharing with the twogenerous puppets, we anticipated that 5-year-olds would show evi-dence of reciprocity by decreasing their degree of self-maximizing,compared to 3-year-olds who should persist in self-maximizingregardless of both puppets’ generous characters. In relation to thecostly punishment task, because both puppets acted with markedgenerosity (hence negating any need to adopt a principled “ethi-cal stance”), we expected a lesser frequency of costly punishmentcompared to Study 2. Furthermore, we predicted that if punish-ment did occur, it would not be oriented systematically towardeither one of the generous puppets.

RESULTSData were analyzed using the same 4 (round) × 3 (recipient) × 2(age) × 2 (gender) mixed-design ANOVA described previously.Because children distributed coins between themselves and iden-tical generous protagonists, the potential recipients for this studyincluded the child, the left-side puppet, and the right-side pup-pet. Gender and children’s performance on the three-way pre-test(included as a covariate) were both non-significant, F 1,19 = 0.479and 0.373, respectively.

Analysis yielded a significant interaction of recipient andage, F 2,18 = 4.83, p = 0.021, η2 = 0.349. Contrasts reveal that asrecipients and with regard to self-maximizing tendencies, 3-year-olds gave more coins to themselves (M ± SD = 4.17 ± 2.02) thandid 5-year-olds (M ± SD = 2.27 ± 1.27), F 1,19 = 5.24, p = 0.034,η2 = 0.216. Three-year-olds also tended to give more coinsto themselves than to either the left (M ± SD = 2.18 ± 1.97)or right (M ± SD = 2.76 ± 1.67) puppets, both p < 0.05 basedon Bonferroni-adjusted pairwise comparisons. Amongst 5-year-olds there was a non-significant trend for children togive more coins to the left (M ± SD = 3.40 ± 1.79) and tothe right puppet (M ± SD = 3.34 ± 1.56) than to themselves(M ± SD = 2.27 ± 1.28), suggesting that these children weresharing with almost absolute equity. Neither 3- nor 5-year-olds exhibited signs of a side bias by preferentially giv-ing significantly more coins to the right versus the left-side puppet (see Figure 4, left and right panels, respec-tively).

www.frontiersin.org December 2011 | Volume 2 | Article 353 | 7

Robbins and Rochat Emerging signs of strong reciprocity

FIGURE 4 | Children’s distributions in triadic sharing in Study 3

(Generous Control), Study 4 (Stingy Control) and Study 5

(Non-Agentive Control). Children split coins between themselves andtwo puppets that were either identically generous (Study 3), or stingy(Study 4), or were passive recipients of the game (Study 5). Vertical axis

represents mean number of coins (maximum 9) distributed by3-year-olds (left panel) and 5-year-olds (right panel) to each recipientcollapsed across round. Note that to control for the possible effect ofinequity aversion in older children, only 5-year-olds were tested inStudy 5.

Note that despite these developmental differences, both 3- and5-year-olds in this Study were on average less self-maximizing thantheir counterparts in Study 2. We computed a “self-maximizing”score by averaging the number of coins given to the child acrossthe four sharing rounds and used this score as the dependent mea-sure in an independent samples t -test to compare self-maximizingacross the two studies. Overall, children in Study 3 (GenerousControl) gave significantly fewer coins (M ± SD = 3.21 ± 1.92) tothemselves than did children in Study 2 (Stingy–Generous TriadicShare, M ± SD = 4.53 ± 1.86), t 88 = 2.95, p < 0.01 (one-tail).

These findings suggest that even when sharing with identicalgenerous protagonists, 3-year-old children persist in their maxi-mization of personal gains; their sharing in this control conditionwas akin to the self-maximizing observed in 3-year-olds in bothStudies 1 and 2. In contrast, and confirming our hypothesis, 5-year-olds showed clear signs of reciprocity with the puppets bysharing more generously and self-maximizing significantly lesscompared to Study 2.

In the absence of any overtly unfair sharing on the part ofthe two puppets, we did not expect children at either age to

engage in costly punishment. Owing to the identical characterof the puppets, if costly punishment occurred, we expected theorientation of such punishment to be at chance. A series ofFisher’s exact tests supported these predictions. Approximatelyhalf of 3- (58%, N = 7) and 5-year-olds (42%, N = 5) engagedin costly punishment; overall, this cohort punished significantlyless frequently than did children in Study 2, Fisher’s exact test:p < 0.01 (one-tail). Furthermore, when children punished theydid so without clearly aligning punishment toward one puppet.Note that 71% of 3-year-olds and 60% of 5-year-olds punishedboth puppets. Linear regression demonstrated that the associa-tion between children’s age (in months) and the percentage ofcoins (out of all coins sacrificed) given to punish the left pup-pet was non-significant, R_1, 102 = 0.02 (Figure 5, center panel).That 5-year-olds show no signs of selective costly punishment isin sharp contrast to the findings reported in Study 2. We spec-ulate that when they punish in this control study, children ofboth age groups act instead of punishing proper to prolong theplayful exchange with the Experimenter (ludic inertia, see GeneralDiscussion).

Frontiers in Psychology | Developmental Psychology December 2011 | Volume 2 | Article 353 | 8

Robbins and Rochat Emerging signs of strong reciprocity

In all, Study 3 confirms that between 3 and 5 years, childrenbecome more sensitive to the generous character of sharing pro-tagonists, aligning with their generous tendencies in acts of reci-procity. Arguably, in the absence of any overt violation of fairnessnorms on the part of the two generous puppets, children were notmotivated to adopt an ethical stance, opting to engage in selectivecostly punishment only at chance levels. For further control, inStudy 4, we repeated the experiment with two stingy puppets.

STUDY 4 (STINGY CONTROL WITH TRIADIC SHARE ANDCOSTLY PUNISHMENT)To assert that children are indeed sensitive to the stingy charac-ter of the puppet, in a second control condition we sampled 24children (12 three-year-olds and 12 five-year-olds, see Table 1)who played the Triadic Sharing Game and Costly Punishment taskdescribed previously with two identically stingy puppets.

HYPOTHESESRegarding the four rounds of three-way sharing with identicalstingy puppets, we expected that if 5-year-olds are engaging in reci-procity, they should increase their tendency to self-maximize,moreclosely resembling 3-year-olds and giving themselves more coinscompared to Study 2. For the costly punishment task, we antic-ipated at both ages no evidence of selective punishment toward

either one of the two puppets owing to the identical nature oftheir characters (i.e., stingy character in both).

RESULTSBecause children distributed coins between themselves and identi-cal stingy protagonists, potential recipients of sharing included thechild, the left puppet, and the right puppet. Triadic share data wereanalyzed using a 4 (round) × 3 (recipient) × 2 (age) × 2 (gender)mixed-design ANOVA. Gender and performance on the three-waypre-test (included as a covariate) were non-significant,F 1,18 = 2.82and 0.565, respectively.

A main effect of recipient (F 2,17 = 55.8, p < 0.01, η2 = 0.112)demonstrates that 3- and 5-year-olds were similar in how theydistributed coins amongst the three protagonists (Figure 4, leftand right panels, respectively). Children at both ages maxi-mized their gains, giving themselves on average 6.5 (roughly72%) of the nine coins. Mean differences indicate that childrengave significantly more coins to themselves than to either theleft (MD ± SE = 5.45 ± 0.402) or right (MD ± SE = 5.42 ± 0.452)puppet, both p < 0.01 based on Bonferroni-adjusted pairwisecomparisons. Neither 3- nor 5-year-olds exhibited signs of a sidebias by preferentially giving significantly more coins to the rightversus the left-side puppet (see Figure 4, left and right panels,respectively).

FIGURE 5 | Coins sacrificed to punish during the costly punishment task

in Studies 2–4. Vertical axis represents the proportion (%) of the total coinssacrificed by each child to punish the stingy puppet (Study 2, left panel) or the

left-side puppet (Study 3, center, and Study 4, right panel) as a function of thechild’s age in months. The horizontal line represents equal punishment towardthe two puppets.

www.frontiersin.org December 2011 | Volume 2 | Article 353 | 9

Robbins and Rochat Emerging signs of strong reciprocity

Children’s degree of self-maximizing in this condition was ele-vated even above levels noted in Study 2. To compare acrossstudies, we first computed a “self-maximization” score by aver-aging together the number of coins distributed to the child acrossthe four sharing rounds. This value was entered as the depen-dent measure in an independent samples t -test. Regardless of age,when in the presence of two stingy protagonists children give sig-nificantly more coins to themselves (M ± SD = 6.52 ± 1.33) thanin situations in which they distribute resources with a stingyand a generous puppet, as in Study 2 (M ± SD = 4.53 ± 1.86),t 88 = 4.83, p < 0.01 (one-tail).

Consistent with the generous control condition of Study 3, both3- and 5-year-olds were at chance regarding the inclination toengage in costly punishment. Overall, children in this cohort weresignificantly less likely to punish than were children in Study 2,Fisher’s exact test: p < 0.01 (one-tail). Furthermore, children whopunished were non-selective in their orientation of punishment:67% of 3-year-olds and 83% of 5-year-olds punished both puppetsequally often. A linear regression model testing the relationshipbetween children’s age (in months) and the percentage (out oftotal sacrificed coins) given to punish the left puppet yielded anon-significant association, R_1, 152 = 0.05 (see Figure 5, rightpanel).

In all, the results confirm our prediction that when sharing withtwo stingy characters, both 3- and 5-year-old children demonstratestrong self-maximizing tendencies. As in Study 3 (generous con-trol) but in contrast to Study 2 (stingy and generous partners),5-year-olds show no sign of selective costly punishment whenagreeing to sacrifice one of their coins at the end of the sharinggame.

To further control that 5-year-olds do construe and factor thestingy and generous characters of the puppets as sharing agents,we performed a final control with non-agentive (passive) puppets.The rationale was that if 5-year-olds were only considering the out-come of the sharing (i.e., unequal coin distribution independentof the agentive character of the puppet), we should replicate thefindings of Study 2. If not, the strong reciprocity (ethical stance)hypothesis would find further support to account for the behaviorof 5-year-olds.

STUDY 5 (NON-AGENTIVE CONTROL WITH TRIADIC SHAREAND COSTLY PUNISHMENT)In a final control we tested the possibility that sharing outcome(e.g., the number of coins in the banks at the end of the Tri-adic Sharing game) rather than the character and agency of thepuppets could account for signs of strong reciprocity. In Study5, children playing the same Triadic Sharing game described pre-viously were presented with a pre-established distribution, neverseeing the puppets actually sharing either stingily or generouslybut resting inanimate in front of the child. Prior to the child’sturn to split the coins, the Experimenter displayed three groups ofcoins on a tray, one for each of the three protagonists. The coinswere presented in the same ratios as in Studies 1and 2. Thus, inone turn the “stingy” puppet received four coins while the childand “generous” puppet received only one each, and in anotherturn the “generous” puppet received one coin while the child and

“stingy” puppet each received four coins. As in the previous stud-ies, we counterbalanced the location of the two puppets relativeto the child as well as the order in which the stingy and gener-ous distributions were presented. When it was the child’s turn,the child actively distributed as in the other studies. At the end ofthe Triadic Sharing game children participated in the same CostlyPunishment task described previously. Because in Study 2 selec-tive costly punishment oriented toward the stingy protagonist wasonly observed in older children, we limited our sample in this newcontrol condition to only 5-year-olds (see Table 1).

HYPOTHESISFollowing the rationale of this control (see above) and based on thestrong reciprocity hypothesis demonstrated in 5-year-olds (Study2), we expected significantly less signs of oriented costly pun-ishment in this non-agentive, passive puppet condition (no clearagentive character of the puppets).

RESULTSOnce again, triadic share data were analyzed using a 4 (round) × 3(recipient) mixed-design ANOVA. On the basis of our earlier find-ings, gender and the pre-test were omitted as factors in analysis.Results yielded a main effect of recipient, F 2,14 = 5.566, p = 0.017,η2 = 0.263. On average children gave significantly more coinsto themselves (M = 4.521, SD = 0.484) than to either the stingy(M = 2.118, SD = 0.313) or generous (M = 2.361, SD = 0.264),both p < 0.05 based on Bonferroni-adjusted pairwise compar-isons. Unlike their counterparts in Study 2, children in this controlcondition were less self-maximizing and did not discriminatebetween the two non-agentive puppets (Figure 4, right panel).

With regards to the costly punishment task, a binomial testrevealed that significantly more children (N = 14, or 78%) optedto punish than not, p = 0.015 (one-tail). This value did not differsignificantly from the 91% of 5-year-olds who opted to punish inStudy 2 based on a Fisher’s exact test.

However, differences emerged with regard to the selectivity ofpunishment. In this control condition, 7 of 14 children (50%)punished both puppets. In contrast, children in Study 2 weremore selective: 17% of children punished both puppets, whereasthe majority (83%) were significantly above chance levels in theirpunishment of only one puppet, p < 0.05. In a more direct Fisher’sexact test of these two studies, our results show a significanttrend by which children in Study 2 were more likely to ori-ent punishment toward only one puppet than were children inthe current control, p < 0.05. However, when exclusive punish-ment of one puppet did occur, results show that, as in Study2, it was oriented toward the stingy protagonist. Of the chil-dren who sacrificed coins to punish only one puppet, 100% ofchildren in Study 2 and 86% of children in the present con-trol oriented this punishment toward the stingy puppet. Thepercentage of coins sacrificed to punish the stingy puppet wasalso similar between Study 2 (M ± SD = 0.827 ± 0.285) and thepresent control (M ± SD = 0.680 ± 0.337), t 42 = 1.334, p = 0.193(two-tailed). In short, significantly more children punished bothpuppets in the control compared to Study 2. However, when chil-dren opted to punish only one puppet, in both studies the majorityof children punished the stingy protagonist.

Frontiers in Psychology | Developmental Psychology December 2011 | Volume 2 | Article 353 | 10

Robbins and Rochat Emerging signs of strong reciprocity

As in Study 2, to determine whether punishment orientedtoward the stingy protagonist could be predicted by the total accu-mulated coins in each of the three banks, we correlated our index ofinequity aversion with the percentage of coins sacrificed to pun-ish the stingy puppet (see Study 2 for formulation). Consistentwith the results of Study 2, this relationship was non-significant,r(14) = 0.023, suggesting that the differing amounts of coins ineach recipient’s bank at the end of the fourth round did not predictselective punishment toward the stingy character.

These results suggest that in the absence of any agency orexplicit reference to the character of the two puppets, childrenwere significantly less inclined to orient their punishment solelyto the stingy or the generous protagonist, in contrast to what wefound in Study 2. Furthermore, in both studies, when childrenopted to punish the Stingy puppet at least once, this tendencywas not predicted by the relative number of coins in each recipi-ent’s bank at the end of the game. We therefore conclude that theinequity aversion hypothesis cannot account for our results.

STUDY 6 (CROSS-CULTURAL COMPARISON OF US ANDSAMOAN 5-YEAR-OLDS)Egalitarianism and responses to unfair or inequitable acts mightbe deeply rooted in human evolutionary history (Fehr et al., 2008;Olson and Spelke, 2008). Such claims call for further cross-culturalcomparisons that would confirm the universality of strong reci-procity as they emerge in human ontogeny. We had the rare oppor-tunity to test a sample of 5- to 6-year-olds living in a small rural andtraditional Polynesian village of Samoa in the South Pacific (seedescription below). Here we present findings for both the triadicshare and costly punishment replicating the procedure outlinedin Study 2. We compared Samoan children with age and gender-matched children from our sample of American 5-year-olds inStudy 2. The rationale was to probe the role of highly contrastedcultural contexts on children’s early expression of strong reci-procity and to assess the extent to which such development mightbe universal.

SAMOAN CHILDREN’S CULTURAL CONTEXTChildren were recruited from the village of Faga (populationapproximately 500) on the East coast of the island of Savai’i(population approximately 50,000) of independent Samoa (pop-ulation approximately 180,000, with a GDP nominal per capitaof $2,608 USD, as opposed to $46,900 in the United States).Samoa is at the heart of Polynesia and still maintains strict ances-tral Polynesian traditions despite successive colonial dominationsby the British, Germans, and New-Zealanders, notwithstanding awidespread conversion to Christian churches of almost all denom-inations starting the middle of the nineteenth Century. Gainingits political independence in 1962, and despite years of Westerninfluence and governance, Samoa has managed to uphold strongtraditions of collective living organized around a highly hierarchi-cal chief (“Mataï”) system that is typical of Polynesian culture. InSamoa, this system is maintained in arguably its purest form withinthe Polynesian context of the South Pacific (Shore, 1982; O’Meara,1990). Samoan culture and societal organization is a good exampleof what is sometimes characterized as“vertical”as opposed to“hor-izontal” or “egalitarian” collectivism (Triandis and Gelfand, 1998).

Children were tested in the village of Faga, which engages inhorticulture, fishing, and marine foraging for survival and as ameans of small cash revenue, aside from remittance money sentfrom family members living and working abroad, primarily inNew Zealand. Before children attend school (at around age threeto five), they typically spend the day with their immediate andextended family, playing outdoors with other young children, orobserving alongside adults as they perform their daily duties. Thereis minimal adult supervision after the child begins to walk; how-ever children tend to be supervised collectively by the adults andolder children in the village. Aside from collective games aroundballs and very few other shared objects, these children develop ina culture that does not emphasize individual possession. Childrentoys are rare in the Samoan context that emphasizes communalproperties and living space, rather than individual possession andprivate quarters.

METHODOLOGICAL CONSIDERATIONS AND PARTICIPANTSWe tried to replicate with Samoan 5- to 6-year-old children themulti-round, triadic sharing game with stingy and generous pup-pets as well as the costly punishment task described in Study 2. InSamoa we used materials that were more familiar to the partici-pants such that children split buttons between two sock puppets,props that were meant to be culturally less eccentric than the pokercoins and the high-quality plush puppets used with the US chil-dren. Consistent with Studies 1–4, children were informed that,like the coins, buttons were valuable placeholders to be redeemedat a makeshift toy store at the game’s conclusion. Figure 6 depictsa child tested in Samoa.

We tested a 14 children (seven male and seven female, seeTable 1). All children were tested by a trained adult native fromthe village who was fluent in Samoan language. Back translationsof the Samoan Experimenter’s instructions to the child were per-formed to ensure consistency and comparability between Samoanand English.

FIGURE 6 | Experimental situation in Samoa. The plush puppets andpoker chips used for the American samples were replaced with moreculturally familiar items (sock puppets and buttons).

www.frontiersin.org December 2011 | Volume 2 | Article 353 | 11

Robbins and Rochat Emerging signs of strong reciprocity

RESULTSFor the triadic sharing game, the number of buttons (outof nine) distributed to each player was analyzed using a 4(round) × 3 (recipient) repeated measures ANOVA. Gender andperformance on the three-way pre-test were not included as fac-tors because there was very little variation in sharing betweenmales and females, and all but 1 child in the Samoan sampleshared coins equitably at pre-test. Similarly, because our sam-ple was limited to 5-year-olds, age was not factored into theanalysis.

We observed a main effect of recipient, F 2,12 = 4.86,p = 0.029, η2 = 0.442. Pairwise comparisons indicate that in gen-eral, unlike American children, Samoan children gave fewercoins to themselves (M ± SD = 2.29 ± 0.282) than to eitherthe stingy puppet (M ± SE = 3.46 ± 0.265) or the generouspuppet (M ± SD = 3.21 ± 0.175), both p < 0.05 (Bonferroni-adjusted). Also in contrast to the American 5-year-oldsof Study 2, our Samoan sample did not discriminatebetween the two puppets, giving roughly the same num-ber of coins to both generous and stingy protagonists,MD ± SE = −0.167 ± 0.241.

With regard to costly punishment, Samoan 5-year-olds optedto punish at chance levels: Eight children (57%) engaged in costlysacrifice whereas six children (43%) did not, both p > 0.05 basedon binomial tests. When they punished, children were also atchance regarding who should be punished, with approximatelyhalf of the children punishing stingy alone, and half punishingboth characters.

DIRECT COMPARISON WITH THE US SAMPLEWe compared our sample of Samoan 5-year-olds (Figure 7, leftpanel) to age and gender-matched, randomly chosen 5-year-oldsfrom our sample of American children in Study 2 (seven malesand seven females, see Table 1).

Regarding the triadic share (see Figure 7), we computeda “self-maximizing” score for our matched sample of Samoanand American 5-year-olds by averaging the number of coinsdistributed to the self across the four sharing rounds. Thisscore was then used in an independent samples t -test. Resultsshow that American children gave themselves significantlymore coins (M ± SD = 4.70 ± 1.55) than did Samoan children(M ± SD = 2.29 ± 0.976), t 26 = 4.55, p < 0.01(one-tailed).

FIGURE 7 | Samoan and American 5-year-olds’ distributions in triadic

sharing in Study 6. Vertical axis represents mean number of coins (maximum9) distributed by Samoan children (left panel) versus a randomly selected

sample of age and gender-matched children from Study 2 (right panel) to eachprotagonist (child, left puppet, right puppet) as a function of round (1–4). Thehorizontal line represents absolute equity (three coins). Bars represent ±1 SE.

Frontiers in Psychology | Developmental Psychology December 2011 | Volume 2 | Article 353 | 12

Robbins and Rochat Emerging signs of strong reciprocity

Samoan 5-year-olds also differed in their treatment of thetwo puppets. We analyzed the number of coins distributedto protagonists in a 4 (round) × 2 (recipient: stingy or gen-erous) × 2 (culture) mixed-design ANOVA, finding a signifi-cant interaction of culture and recipient, F 1,26 = 8.31, p = 0.027,η2 = 0.225. Across the four sharing rounds, American childrenpreferentially shared with the generous over the stingy pup-pet, MD ± SE = 0.492 ± 0.176, p < 0.01 (based on Bonferroni-adjusted pairwise comparisons). In contrast, although Samoanchildren gave slightly more coins to the stingy puppet, the meandifference in the number of coins distributed to the puppets wasnon-significant, MD ± SE = −0.167 ± 0.241. Samoan childrendid not preferentially share with either puppet.

With regard to costly punishment, a Fisher’s exact test reveals anon-significant trend of more American children (86%, N = 12)opting to punish compared to 57% (N = 8) of Samoan children.When children agreed to give up coins to punish, the percent-age sacrificed to punish the stingy puppet did not differ signif-icantly across cultures, M ± SD = 0.817 ± 0.333 for the US andM ± SD = 0.576 ± 0.348 for Samoa, based on an independentt -test: t 18 = 1.48, p > 0.05. However, when examining each cul-ture independently, only US children were above chance levels(p = 0.005) in choosing to punish, based on binomial tests. Thesefinding suggests that although punishment may be less commonin Samoa, when punishment does occur, like in the US, it tendsto be oriented more toward the stingy puppet. However, notethat because of the small sample size, there might be insufficientpower to detect a statistically significant difference between thetwo cultures. Alternatively, Samoan children may be less inclinedto punish, but also less inclined to engage in strong reciprocity ori-ented toward a protagonist who has violated an expected norm offairness. It is possible that norms of fairness might differ betweenSamoan and US children. More research is necessary to furtherassess the relative universality of strong reciprocity, with a focuson the relationship between cultural context and the inclinationto selectively punish. This said, our preliminary investigation con-firms that there might be important, yet subtle cross-culturaldifferences.

GENERAL DISCUSSIONCollectively, our findings provide first evidence that what we con-strue as strong reciprocity emerges by 5 years in human ontogeny,at least in middle class US children. Although 3-year-olds mod-ulate their tendency toward self-maximizing depending on therelative generosity or stinginess of sharing partners, this early sen-sitivity is not yet principled. In 3-year-olds, there are no apparentsigns of selective costly punishment. Our research suggests thatsigns of strong reciprocity emerge from 5 years of age.

Furthermore, our analysis shows that a simple aversion toinequity cannot account for what we observe. If some sensitiv-ity to inequity is necessary for the expression strong reciprocity, isnot sufficient. In the expression of strong reciprocity as construedhere, individuals not only detect but also tend to act principallyupon it by either punishing or rewarding, even if doing so comesat a personal cost. The developing propensity to act, even at apersonal cost, captures first signs of an ethical, morally principledstance emerging by 5 years of age and not earlier.

The results of the triadic sharing game in Studies 1 and 2 suggestthat although children in both age groups express self-maximizingtendencies, only 5-year-olds demonstrate evidence of reciprocityby discriminating between the generous and stingy puppets, opt-ing to give more coins to the generous puppet and reducing payoffsto the stingy puppet over the four rounds of sharing. In the twofollow-up control conditions, 5-year-olds showed further signsof reciprocity by decreasing their self-maximizing behavior whensharing with identical generous puppets (Study 3) and increas-ing this self-maximization when interacting with identical stingypuppets (Study 4). When sharing with non-agentive characters(Study 5), 5-year-olds self-maximized their payoffs (though to alesser extent than in Studies 1 and 2) and did not discriminatesignificantly between the“stingy”and“generous”protagonists whowere passive actors and recipients in the game.

Study 2 provides further evidence that although children ofboth age groups engage in costly punishment, only 5-year-oldsshow signs of strong reciprocity by orienting their costly sacri-fice markedly more toward the stingy protagonist. Three-year-oldchildren do not align punishment with either character, engagingin indiscriminate punishment that may be more about contin-uing the inertia of the game rather than conveying approval ordisapproval for the puppets’ actions. Selective costly punishmentwas only observed in conditions where children interacted withboth an agentive stingy and an agentive generous protagonist. Incontrol conditions where puppets were either identically generous(Study 3) or identically stingy (Study 4), both 3- and 5-year-oldswere at chance with regard to the inclination to punish, and toorient punishment toward a particular protagonist. In the thirdcontrol condition (Study 5), when the puppets were non-agentiveand the distribution pre-established, 5-year-olds were significantlyless inclined to punish only one of the two puppets, whether stingyor generous. In thus appears that by 5 years, children show signif-icantly more signs of strong reciprocity. Our analyses also showthat such signs of strong reciprocity cannot be merely reduced toan expression of inequity aversion. By 5 years, the costly behav-ior of the tested children cannot be predicted by the unequalaccumulation of coins at the end of the triadic share. Instead,the inclination to engage in costly sacrifice seems to depend onthe relative character of the child’s sharing partners (i.e., stingyversus generous), particularly so when these partners behave asagents.

With regard to the relative universality and cross-culturalequivalence in the expression of strong reciprocity in ontogeny,a comparison of age-matched Samoan and American children(Study 6) suggests that the proclivity to engage in costly pun-ishment is at 5 years of age more prevalent in the US com-pared to Samoa. Although further research is needed, our datasuggest that culture could play a role in the expression of anethical stance in child development. The collectivist, more group-oriented culture surrounding Samoan compared to US children,and particularly compared to the middle class US sample testedhere, might impact their developing sense of fairness expressedin distributive justice. Samoan 5-year-old children did show aheightened tendency toward egalitarianism and lesser propen-sity toward self-maximizing in the context of our triadic sharinggame.

www.frontiersin.org December 2011 | Volume 2 | Article 353 | 13

Robbins and Rochat Emerging signs of strong reciprocity

Questions remain as to what factors contribute to such devel-opment. Recent research suggests that concern for others in dis-tributive justice games may be determined in part by parochialismand close relations (Fehr et al., 2008; Olson and Spelke, 2008).Favoritism toward perceived in-group members becomes a majorpredictors of fair distributions, particularly in children 5 years andolder. The relationship between the development such distribu-tive behavior and so-called moral emotions (e.g., spite, altruism)would need to be further investigated. Finally, evidence of third-party punishment (e.g., costly sacrifice to punish one who haswronged another) in 5-year-old children could strengthen theyethical stance hypothesis proposed here.

CONCLUSIONBeyond the speculation that humans might be born with a unique,species-specific “moral instinct,” our own research points towardthe importance of examining the role of culture in the expres-sion of such fair-minded behavior and the child’s ability to adoptan ethical stance. Preliminary cross-cultural differences reportedhere, although limited, point to various possibilities. For example,Samoan and US middle class cultures might be associated withdifferent developmental trajectories regarding the emergence of

strong reciprocity. Alternatively, children of both cultures mightexpress strong reciprocity at about the same age, but in differ-ent ways reflecting, for example, their relative inclination towardsocial conformity, their relative respect for rules and adult author-ity (Keller et al., 2004), as well as their greater sense of fairnessand aversion toward inequality (Fehr et al., 2008; Rochat et al.,2009).

We conclude that strong reciprocity as the putative force behindhuman cooperation, as well as the human tendency to abide byand enforce social norms, might first emerge by 5 years of age andmay depend on culture. From then on, children begin to situatethemselves, their understandings, judgments, and actions towardothers in a morally principled space (Taylor, 1989). This, we pro-pose, represents a radical ontogenetic shift that is deserving offurther empirical scrutiny.

ACKNOWLEDGMENTSOur gratitude goes to Britt Berg and the students of the EmoryUniversity Infant and Child Lab for data collection, coding, andanalysis. Our thanks also go to Tanya and James Broesch for help-ing in the conceptualization of the project in its early stages andfor their assistance collecting data in Samoa.

REFERENCESCallaghan, T., Rochat, P., Lillard, A.,

Claux, M. L., Odden, H., Itakura, S.,Tapanya, S., and Singh, S. (2005).Synchrony in the onset of mental-state reasoning: evidence from fivecultures. Psychol. Sci. 16, 378–384.

Damon, W. (1975). Early conceptionsof positive justice as related to thedevelopment of logical operations.Child Dev. 46, 201–312.

Dupoux, E., and Jacob, P. (2007). Uni-versal moral grammar: a criticalappraisal. Trends Cogn. Sci. (Regul.Ed.) 11, 373–378.

Eisenberg, N., and Fabes, R. A. (1998).“Prosocial behavior and develop-ment,” in Handbook of Child Psy-chology, ed. B. Damon (New York:Academic Press), 701–778.

Faigenbaum, G. (2005). Children’s Eco-nomic Experience: Exchange, Reci-procity, and Value. Buenos Aires:Libros En Red.

Fehr, E., Bernhard, H., and Rockenbach,B. (2008). Egalitarianism in youngchildren. Nature 454, 1079–1083.

Fehr, E., Fischbacher, U., and Gâchter,S. (2002). Strong reciprocity, humancooperation, and the enforcementof social norms. Hum. Nat. 13,1–25.

Haidt, J., and Joseph, C. (2004). Intuitiveethics: how innately prepared intu-itions generate culturally variablevirtues. Daedalus 133, 55–66.

Hamlin, J. K., Wynn, K., and Bloom, P.(2007). Social evaluation by prever-bal infants. Nature 450, 557–560.

Hauser, M. (2006). Moral Minds. NewYork: Ecco.

Henrich, J., McElreath, R., Barr, A., Ens-minger, J., Barrett, C., Bolyanatz,A., Cardenas, J. C., Gurven, M.,Gwako, E., Henrich, N., Lesoro-gol, C., Marlowe, F., Tracer, D., andZiker, J. (2006). Costly punishmentacross human societies. Science 312,1767–1770.

Henrich, N., and Henrich, J. (2007).Why Humans Cooperate: A Cul-tural and Evolutionary Explana-tion. NY, Oxford: Oxford UniversityPress.

Hook, J. (1978). The developmentof equity and logico-mathematicalthinking. Child Dev. 49, 1035–1044.

Inhelder, B., and Piaget, J. (1964). TheEarly Growth of Logic in the Child.New York: Norton.

Kagan, J. (1988). Tempermental con-tributions to social behavior. Am.Psychol. 44, 668–674.

Keller, H., Yovsi, R., Borke, J., Kart-ner, J., Henning, J., and Zaira,P. (2004). Developmental conse-quences of early parenting expe-riences: self-recognition and self-regulation in three cultural commu-nities. Child Dev. 75, 1745–1760.

Kohlberg, L. (1981). Essays on MoralDevelopment. San Francisco: Harper& Row.

Kuhlmeier, V. A., Wynn, K., and Bloom,P. (2003). Attribution of disposi-tional states by 12-month-olds. Psy-chol. Sci. 14, 402–408.

Lane, I. M., and Coon, R. C. (1972).Reward allocation in preschool chil-dren. Child Dev. 43, 1382–1389.

McCrink, K., Bloom, P., and Santos,L. R. (2009). Children’s and adults

judgments of equitable resource dis-tributions. Dev. Sci. 13, 37–45.

Olson, K. R., and Spelke, E. S.(2008). Foundations of cooperationin young children. Cognition 108,222–231.

O’Meara, T. (1990). SamoanPlanters: Tradition and EconomicDevelopment in Polynesia. FortWorth: Holt, Rinehart, & Winston,Inc.

Piaget, J. (1932). The Moral Judgmentof the Child. London: Routledge &Kegan Paul.

Pinker, S. (2008). The Moral Instinct.New York Times Magazine, January12.

Rochat, P., Dias, M. D. G., Guo, L.,Broesch, T., Passos-Ferreira, C.,Win-ning, A., and Berg, B. (2009). Fair-ness in distributive justice by 3- and5-year-olds across 7 cultures. J. CrossCult. Psychol. 40, 416–442.

Shore, B. (1982). Sala’ilua: A SamoanMystery. New York: Columbia Uni-versity Press.

Snarey, J. (1985). Cross-cultural uni-versality of social-moral develop-ment: a critical review of Kohlber-gian research. Psychol. Bull. 97,202–232.

Taylor, C. (1989). The Sources of the Self:The Making of the Modern Identity.Cambridge, MA: Harvard UniversityPress.

Tomassello, M. (2008). The origins ofcommunication. Cambridge, MA:MIT Press.

Triandis, H. C., and Gelfand, M. J.(1998). Converging measurement ofhorizontal and vertical individual-

ism and collectivism. J. Pers. Soc.Psychol. 74, 118–128.

Warneken, F., Hare, B., Melis, A.P., Hanus, D., and Tomasello, M.(2007). Spontaneous altruism bychimpanzees and young children.PLoS Biol. 5, e184. doi:10.1371/jour-nal.pbio.0050184

Wellman, H. M., and Liu, D. (2004).Scaling of theory-of-mind tasks.Child Dev. 75, 523–541.

Zahn-Waxler, C., Radke-Yarrow, M.,Wagner, E., and Chapman, M.(1992). Development of concern forothers. Dev. Psychol. 28, 126–136.

Conflict of Interest Statement: Theauthors declare that the research wasconducted in the absence of anycommercial or financial relationshipsthat could be construed as a potentialconflict of interest.

Received: 27 August 2011; accepted: 09November 2011; published online: 19December 2011.Citation: Robbins E and Rochat P (2011)Emerging signs of strong reciprocityin human ontogeny. Front. Psychology2:353. doi: 10.3389/fpsyg.2011.00353This article was submitted to Frontiers inDevelopmental Psychology, a specialty ofFrontiers in Psychology.Copyright © 2011 Robbins and Rochat.This is an open-access article distributedunder the terms of the Creative CommonsAttribution Non Commercial License,which permits non-commercial use, dis-tribution, and reproduction in otherforums, provided the original authors andsource are credited.

Frontiers in Psychology | Developmental Psychology December 2011 | Volume 2 | Article 353 | 14